Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror

Comment Re: Wayland mostly works for me (Score 1) 131

We have >60k users using remote access. They all seem fine with it.

I've never noticed any real latency difference between ssh, RDP, VNC, or the like unless my OWN local Internet connection is starved for bandwidth. Compressed RDP and VNC are relatively small, like 1-3Mbps with a 4k-ish display. It can sometimes peak up to 6-12Mbps because there's some stupid full screen video.

Comment Re:you know why? (Score 1) 44

I've yet to see Kotlin outside of Android Apps. I've heard some Fintech startups were using it, but I have not seen it despite my job being Fintech consulting. What I do see is Clojure.

I've seen Swift backends in production serving millions of people (billions even.)

Neither is a particularly difficult to learn Language and TBH, I suspect anyone who knows one can learn the other with minimal investment.

Comment Re: Near native performance? (Score 1) 29

They didn't fuck up. This is not a full sized disk format aka my VM's "128GB disk" is not a 128GB+ sized file on the native file system, if it was it would be mostly close to native (still not native because of indeterminate caching and user space vs "kernel" space access issues.) This is basically approaching raw disk performance.

Comment Re: Near native performance? (Score 1) 29

No "thin" provisioned disk format is as fast native, it's not even especially close in many circumstances. The "thick" provisioned disk formats are generally much closer, similar to this one. This is definitely a very fast thin format.

Thin = roughly the size of the data or at least the data marked blocks.
Thick = roughly the full size of the disk volume (sometimes they are compressed or sometimes they are a bit larger than "full size.)

Comment Re:Oh Apple (Score 1) 60

We had this legacy Solaris on SPARC service, 3 layer, sounds similar architecture but more servers, cores, fairly good amount of storage. Recently we had issues with Oracle pegging CPU cores and there's a 2 year late project to replace it with almost 200 servers (because growth and shit.)

I replaced it with 2 containers (for the App itself) running on a 2 VM cluster mounting from our ceph cluster (standard enterprise service) and a containerized postgresql cluster (standard enterprise service). Two week hack-a-thon with 3 developers and myself. We did rewrite the application from ancient C++ to Go.

The App containers are expandable, but in the 60 days in production we haven't had to expand them. The postgresql cluster didn't have any noticeable increase and same with ceph cluster. We did discuss doing a microservices architecture, but that only makes sense in certain circumstances and this was not it.

Comment Re:but did they... (Score 3, Interesting) 101

There's some parts of brains that mostly look and operate similar to LLM and the question is, do those have some built-in mechanism to deal with LLM Hallucinations? (Or is the difference in operations enough that it doesn't occur in the electro-organic version.)

We have some ideas of why they hallucinate, it's not like we're flying blind.

Slashdot Top Deals

Center meeting at 4pm in 2C-543.

Working...